Your next prospect already asked ChatGPT to shortlist agencies before they ever found your site. Here's the AI citation audit that finds out if you made the cut.
by Ayush Gupta's AI
The problem
Forrester's 2026 buyer survey found 94% of B2B buyers used AI during their most recent purchase, 55% used it to compare vendors directly, and 95% of deals go to whoever made the buyer's Day One shortlist, a shortlist formed independently before any sales contact. Over half of B2B brands surveyed have zero citations across ChatGPT, Perplexity, and Gemini. Most agency founders have never once typed the query a prospect would type — 'best SEO agency for a Series A SaaS company,' 'top web dev shop for e-commerce replatforms' — into an AI assistant to see who gets named. If the agency isn't in that answer, it's not losing the pitch. It's not getting invited to one.
The fix
Run the actual category queries a prospect would ask across ChatGPT, Perplexity, Gemini, and Claude, log whether the agency gets cited and from what sources, diagnose the specific content gaps causing the miss, and close them with the citation-worthy proof-of-work content and third-party mentions LLMs actually pull from.
The Playbook
Run the queries a prospect would actually type, not your brand name
Searching your own agency name and finding a clean answer means nothing — that's not how buyers start. Write out 8-10 category queries a real prospect would ask: service plus niche, service plus city, service plus company stage. Run each one across ChatGPT, Perplexity, Gemini, and Claude in fresh sessions, and log the raw answer, who got named, in what order, and which sources each tool cited to back up the recommendation.
You are helping me build a category-query test list for an AI shortlist visibility audit.
My agency does: [SERVICE, e.g. "SEO for B2B SaaS companies"]
Our typical client looks like: [CLIENT PROFILE]
Our real competitors are: [COMPETITOR NAMES]
Generate 8-10 realistic queries a prospect would type into ChatGPT or Perplexity while researching agencies in this category, ranging from broad ("best [service] agency for [niche]") to specific (stage, budget, region, use case). Do not include our brand name in any query. Format as a plain numbered list ready to paste one at a time into each AI tool.Diagnose why the agency is missing, not just that it's missing
A blank result is a symptom, not a diagnosis. Feed the raw transcripts back into Claude and ask it to reason about the citation pattern: are competitors getting named because of directory listings, press mentions, case studies with hard numbers, or third-party reviews the agency doesn't have an equivalent of? The fix only works if it targets the actual gap instead of guessing.
I ran these category queries across ChatGPT, Perplexity, and Gemini and I'm pasting the raw answers below. Our agency was not named, or was named without detail.
Queries and raw AI answers:
[PASTE TRANSCRIPTS]
Analyze:
1. Which competitors got cited repeatedly, and what kind of source is each citation likely pulling from (case study, directory, press, review site, own blog)?
2. What content type or proof point shows up in competitor citations that our site likely doesn't have in equivalent form?
3. Rank the 3 most likely reasons we're missing from these answers, most fixable first.Turn finished work into citation-ready proof, not just portfolio pages
LLMs quote specifics: named client type, concrete numbers, a clear before-and-after, dated and attributable. A portfolio page that says 'we helped a SaaS company grow' gives a model nothing to cite with confidence. Rewrite the strongest 3-5 case studies into a structured format an AI tool can lift directly into an answer.
Turn this project summary into a citation-ready case study structured for AI visibility, not just human readers.
Project details:
[PASTE ROUGH NOTES: client type, starting point, what we did, result with numbers, timeframe]
Output format:
- One-sentence summary naming the client type, the problem, and the numeric result (this is the sentence a model would quote)
- Three supporting facts, each with a specific number or timeframe
- One sentence naming exactly what we did that a generic competitor wouldn't have thought to do
Keep every claim specific and verifiable. No vague adjectives.Get corroborated outside your own domain, not just louder on it
AI answer engines weight cross-source agreement heavily — a claim repeated only on the agency's own site reads differently than the same claim showing up on a directory, a review platform, and a partner's site too. Prioritize getting the rewritten case studies and a clear service description onto Clutch, G2, industry directories, and any partner or press placement realistically available, instead of only polishing the homepage.
Re-run the audit on a fixed cadence and track it as a real metric
Citation patterns shift as models retrain and competitors publish more. Re-run the same query list quarterly, track whether the agency's citation rate and position are improving, and treat 'do we show up in the AI shortlist' as a standing pipeline metric sitting next to website traffic and lead volume, not a one-time curiosity check.
What changes
The agency knows exactly where it stands in the AI-driven shortlist prospects are building before any sales conversation, closes the specific content gaps causing the miss, and starts showing up in the 95% of deals that get decided before a human ever picks up the phone.
Ask most agency founders how their pipeline looks and they'll talk about SEO rankings, referrals, or LinkedIn outbound. Ask them what ChatGPT says when someone asks it to recommend an agency like theirs, and most have never checked.
That gap is now the pipeline problem that matters most.
The buyer's journey already happened before you knew about it
Forrester's 2026 buyer survey put a number on something a lot of founders sensed but couldn't prove: 94% of B2B buyers used AI during their most recent purchase, 55% used it specifically to compare vendors, and 47% built their internal business case before contacting a single vendor. Most strikingly, 95% of deals go to whoever made the buyer's Day One shortlist — the list formed independently, inside an AI conversation, before sales ever entered the picture.
Over half of B2B brands surveyed have zero citations across ChatGPT, Perplexity, and Gemini. Not weak citations. Zero.
Why "we rank well on Google" doesn't answer this
Traditional SEO and AI answer-engine visibility are related but not the same game. A model isn't ranking ten blue links, it's synthesizing an answer and deciding who's worth naming by weight of evidence: specific case studies with real numbers, cross-source corroboration on directories and review sites, and content structured clearly enough to quote with confidence. An agency can rank fine on Google and still be invisible the moment a prospect asks an AI tool the same question in natural language.
The fix starts with running the actual query
Most agencies have never typed the exact question a prospect would ask into ChatGPT or Perplexity. Not the brand name — the category query: "best [service] agency for [niche]," "[service] shop for a company at [stage]." Running that query in a fresh session and logging who gets named, in what order, and from what sources is the entire audit. It takes twenty minutes and most founders have never done it once.
The gap is almost always the same shape
When an agency comes up empty, it's rarely because competitors are better. It's because competitors have citation-worthy material sitting in places an AI tool actually pulls from: case studies with hard numbers instead of vague adjectives, listings on Clutch or industry directories, press mentions, third-party reviews. The fix isn't a rebrand. It's turning finished client work into specific, quotable, cross-corroborated proof, published somewhere other than just the agency's own homepage.
Bottom line
The AI shortlist is where the deal gets decided now, and it's forming whether or not the agency shows up in it. Run the query, see what the model actually says, and fix the specific gap instead of guessing. The alternative is losing deals to agencies that were never better, just more citable.